[GEODE-7184] Add function execution timers

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.11.0
• Assignee: Aaron Lindsey
• Reporter: Aaron Lindsey
• Created: 10/Sep/19 20:20
• Updated: 30/Dec/19 18:51
• Resolved: 19/Dec/19 23:54

Description

Developers oftentimes deploy their own functions to the system to enable decorator pattern for caching to add information to specific key/value pairs. In doing so, they can introduce bottlenecks into the system where server-side functions can cause issues or make things slower than intended. We want a way that users can view functions that they create, and see what the average execution time looks like.

• Meter Type: Timer
• Name: geode.function.executions
• Description: TBD
• Tags: <common_tags>, function (getId on function, if DNE present getClass.getname of deployed function), succeeded (true/false)

Acceptance Criteria

Meter creation/deletion: Create meter on function execution
Measurement: On an individual server, start the timer when a USER function is invoked/executed, and stop the timer when the user function completes OR errors. If it throws a Function Execution or another error then the tag function.isSuccessful=false

Details on Functions and their execution: https://geode.apache.org/docs/guide/110/developing/function_exec/function_execution.html

Scenarios

Scenario: The timers are created when the function is first executed
Given a user executed a function with ID functionToTime on a cluster with 1 locator/1 server
And functionToTime has not been executed previously
Then the server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = true
• count > 1
• totalTime >= 5,000,000,000ns

And the server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = false
• count = 0
• totalTime = 0

Scenario: Successful singular function execution (registered execution)
Given a user registers a function with ID functionToTime (that waits for 5 seconds) on a cluster with 1 locator/1 server
When functionToTime is triggered using gfsh command: "execute function --id=functionToTime"
And the function completes without error
Then the server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = true
• count = 1
• totalTime >= 5,000,000,000ns

Scenario: Successful singular function execution (unregistered execution)
Given an unregistered function with ID functionToTime (that waits for 5 seconds) exists
When triggered on a client using "FunctionService.onServers(cache).execute(new FunctionToTime())"
And the function completes without error
Then the server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = true
• count = 1
• totalTime >= 5,000,000,000ns

Scenario: Singular function execution with Any Exception
Given an unregistered function with ID functionToTime (that waits for 5 seconds) exists
When triggered on a client using "FunctionService.onServers(cache).execute(new FunctionToTime())"
And the function exits with a Any exception error after running for 5 seconds
Then the server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = false
• count = 1
• totalTime >= 5,000,000,000ns

Scenario: Function execution onRegion multi-server
Given a cluster with 1 locator (named L1) as well as 2 servers (named S1,S2)
And a region called RR1 that is a replicate region
When a function execution is triggered against that replicate region using "FunctionService.onRegion(regionRR1).execute(new FunctionToTime())"
Then one server has the following timer:

• name: geode.function.executions
• tag: id = functionToTime
• tag: succeeded = true
• count = 1
• totalTime >= 5,000,000,000ns

And the other server has the following timer:

• name: geode.cache.function.executions
• tag: id = functionToTime
• tag: succeeded = true
• count = 0
• totalTime = 0

Scenario: Function execution onRegion with partition region multiple times
Scenario: Function execution onRegion multi-server
Given a cluster with 1 locator (named L1) as well as 2 servers (named S1,S2)
And a partition region called PR1 that only exists on S1
When a function execution is triggered 10 times against that replicate region using "FunctionService.onRegion(regionPR1).execute(new FunctionToTime())"
Then S1 has the following timer:

• name: geode.function.executions
• tag:id = functionToTime
• tag:succeeded = true
• count = 10

And S2 has the following timer:

• name: geode.cache.function.executions
• tag:id = functionToTime
• tag:succeeded = true
• count = 0

Scenario: Function execution onRegion with replicate region multiple times
Scenario: Function execution onRegion multi-server
Given a cluster with 1 locator (named L1) as well as 2 servers (named S1,S2)
And a replicate region called RR1 exists
When a function execution is triggered 10 times against that replicate region using "FunctionService.onRegion(regionRR1).execute(new FunctionToTime())"
Then when you aggregate both S1 and S2 server metrics for geode.function.executions with id:functionToTime with succeeded:true then the total count of the aggregate will be 10

Issue Links

• links to: GitHub Pull Request #4038
• links to: GitHub Pull Request #4135
• links to: GitHub Pull Request #4192